The Role of Chromatin Remodeling in Medulloblastoma

Abstract

The unexpectedly high frequency and universality of alterations to the chromatin machinery is one of the most striking themes emerging from the current deluge of cancer genomics data. Medulloblastoma (MB), a malignant pediatric brain tumor, is no exception to this trend, with a wealth of recent studies indicating multiple alterations at all levels of chromatin processing. MB is typically now regarded as being composed of four major molecular entities (WNT, SHH, Group 3 and Group 4), which vary in their clinical and biological characteristics. Similarities and differences across these subgroups are also reflected in the specific chromatin modifiers that are found to be altered in each group, and each new cancer genome sequence or microarray profile is adding to this important knowledge base. These data are fundamentally changing our understanding of tumor developmental pathways, not just for MB but also for cancer as a whole. They also provide a new class of targets for the development of rational, personalized therapeutic approaches. The mechanisms by which these chromatin remodelers are dysregulated in MB, and the consequences both for future basic research and for translation to the clinic, will be examined here.

Introduction

Brain tumors currently account for more than a quarter of all deaths from pediatric cancer 3. Medulloblastoma (MB), the most common embryonal brain tumor, is a major contributor to this unfortunate figure. Improvements in patient outcomes over the past few decades have largely been achieved through advances in the application of chemotherapy and craniospinal radiation, as opposed to the biology‐driven risk stratification that has been so successful in treating leukemia, for example 58. A significant fraction of patients will still die from their disease, and those who survive often experience long‐term side effects 19, 42, 63. Thus, there is clearly scope for a shift toward more rational patient stratification and for the application of personalized targeted therapy based on sound tumor‐biological data 55.

In the recent era of high‐throughput genomics, several major advances have been made in terms of our understanding of the biology behind MB tumorigenesis 47. Arguably, the most important paradigm shift has been the recognition that MB is not a single entity, but rather is composed of multiple molecular subgroups (WNT, SHH, Group 3 and Group 4) that differ in their age and gender distribution, clinical outcome and molecular genetic features 10, 34, 35, 48, 75, 76. Despite these differences, some common threads have also emerged. One such feature that is found across subgroups is the recurrent alteration of the chromatin remodeling machinery, as discussed in the following paragraphs. Even here, however, the importance of molecular classification is highlighted by the fact that certain types of chromatin modifier alterations are restricted to specific subgroups.

The packaging and organization of chromatin is the principal mechanism by which cell‐specific transcriptional programs are established and maintained during organismal development 8, 39, 41, including the specification of neural cell types 28, 73. This organization is mediated in large part by an array of post‐translational modifications (PTMs) at key residues on histone tails—the so‐called histone code 29, 36. Genes involved in this process of chromatin regulation are typically grouped into four main categories. First, there are “writers” of covalent histone marks such as histone methyltransferases (HMTs) and histone acetyltransferases (HATs) that establish patterns of PTMs on histone tails. Next, there are “erasers,” histone demethylases (HDMs) and histone deacetylases (HDACs), which can remove these PTMs. An additional class of genes, the “readers” of PTMs, binds to specific histone tail motifs in order to coordinate the downstream consequences of these marks. Finally, there are also large‐scale remodelers of nucleosome structure and chromatin architecture such as the SWI/SNF, INO80 or CHD/NuRD complexes 11, 25, 38, 61, 78, or CTCF 52.

Disruptions of these normal regulatory processes, including in nervous system formation, have been linked to multiple disorders 23, 37, 64. Indeed, it has become increasingly clear in recent years that alterations in how chromatin structure controls cellular processes are one of the most important features of human cancer, irrespective of tumor site or histology 9, 12, 84. Intriguingly, mutations at key regulatory residues in histone proteins themselves (histone 3.1 and the variant 3.3), as well as the histone chaperones ATRX and DAXX, have recently been identified in a subset of another malignant pediatric brain tumor—glioblastoma 33, 68, 81 (see also Fontebasso et al in this issue). Importantly, these alterations in chromatin modifying genes also provide new classes of candidates for targeted therapeutic intervention, with several compounds already showing promise in preclinical or clinical settings 13, 56.

The purpose of this article is to highlight what is currently known about the role of dysregulation of chromatin remodeling in MB and to provide an outlook as to how this knowledge may be expanded upon in the future, as focus shifts toward translating these findings into benefit for patients.

Copy‐Number Changes and Altered Chromatin in MB

The first strong evidence for a role of altered chromatin in MB development emerged from genome‐wide investigation of DNA copy‐number alterations 49. A study by Northcott et al identified multiple changes affecting chromatin modifying genes, including focal gain or amplification of the histone acetyltransferase KAT6A (formerly MYST3) and the lysine demethylases JMJD2B and JMJD2C, together with deletions of polycomb group genes (L3MBTL2, L3MBTL3 and SCML2) and histone methyltransferases (EHMT1, SMYD4). Many of these genes converge on regulation of H3K9 methylation status, and hypomethylation at H3K9 was revealed by immunohistochemical analysis to occur in 41% of tumor samples investigated 49. The same study demonstrated that re‐expression of L3MBTL3 in DAOY cells, in which the gene is deleted, could restore H3K9 methylation and block cellular proliferation.

This work was recently expanded upon in a comprehensive copy‐number profiling analysis of more than 1000 MBs by an international consortium led by the same Toronto group 50, which confirmed the importance of copy‐number alterations at chromatin modifying genes in a subset of cases. These alterations were particularly enriched in Group 4 MBs, and novel homozygous deletions on chromosome X affecting the H3K27 demethylase KDM6A (also called UTX) were identified in this subgroup. Interestingly, this gene was also found to be recurrently mutated in the same subgroup, as described in the following paragraphs.

It is also interesting to note that alterations of chromatin architecture may be a cause of certain structural changes in a cell's DNA rather than simply a consequence. For example, a process of catastrophic rearrangement occurring at a single time point [termed chromothripsis 70] has recently been linked to TP53‐mutated SHH‐subgroup MB (SHH‐MB) 60. One hypothesis as to how these dramatic rearrangements might arise is through a critical loss of specialized chromatin structures at the telomeres, resulting in chromosome end‐to‐end fusions and subsequent mechanical shearing during mitosis 77. A link between germline TP53 mutation [as was frequently observed in SHH‐MB with chromothripsis 60], accelerated telomere shortening and age of tumor onset has previously been established 72. It will therefore be of interest to further investigate the role of telomeric chromatin alterations in the generation of these complex DNA copy‐number changes.

Mutations of Chromatin Modifying Genes—Insights from Large‐Scale Sequencing Studies

The first large‐scale, unbiased sequencing effort in MB, published by Parsons et al in 2011, provided a number of further insights into the role of chromatin modifying genes in this tumor type 53. Several truncating mutations in the histone methyltransferases MLL2 and MLL3 were reported, suggesting a tumor suppressor function of these two genes in MB. Alterations in chromatin modifiers in general were clearly a recurring theme, with rarer mutations being found in the histone demethylase KDM6B, as well as the SWI/SNF chromatin remodeling complex members SMARCA4 and ARID1A, among others. The subgroup specificity of these changes, however, was not clear in this study.

The theme of chromatin modification was also an obvious feature of three next‐generation sequencing studies published in the summer of 2012 30, 57, 62, and was summarized in a recent review and meta‐analysis 47. One‐third of all tumors, across all subgroups, were reported as having a mutation in a gene mapping to the Gene Ontology (GO) term “Chromatin modification” (GO:0015168). A summary of recurrently mutated chromatin modifiers, and their frequency of mutation, is shown in Figure 1.

Figure 1.

Chromatin modifier genes recurrently altered in medulloblastoma (MB). This figure indicates those genes related to the GO term “Chromatin Modification” (GO:0015168), which are mutated at least twice in the discovery cohorts (whole‐exome or whole‐genome sequencing, n = 189) of the three next‐generation sequencing studies described herein 30, 57, 62. Percentages by gene names indicate the overall frequency of mutation in this gene across the three studies. NB—cases with no mutation in the listed genes are not shown.

Mutations of SMARCA4, as identified in the Parsons et al study, were found to be largely restricted to WNT and Group 3 MB, where they were identified in 25% and 11%, respectively, of tumors in these subgroups 47. Rarer mutations in ARID1B and ARID2 were also observed, again implicating the SWI/SNF complex as an important component of MB development. The question of whether there may be a link between hyperactivation of c‐Myc oncogene signaling (common to WNT and Group 3 tumors) and alterations in this complex will be an interesting avenue for further investigation.

SHH‐MBs showed a strong enrichment for alterations in several members of the nuclear receptor co‐repressor (N‐CoR) complex, which is associated with histone deacetylation and is thought to repress target genes by inducing chromatin condensation 51. Genes of this complex mutated in SHH‐MB include BCL6 co‐repressor (BCOR), LIM‐domain binding 1 (LDB1) and G‐protein pathway suppressor 2 (GPS2). In addition, LIM‐domain only 4 (LMO4), a binding partner of LDB1 14, has been identified as a recurrently amplified gene in SHH‐MB 50, further supporting a key role for this complex in the pathogenesis of this subgroup.

One of the most striking single gene changes was the recurrent inactivating mutation of KDM6A in 12% of Group 4 tumors 47. This gene acts in concert with the most commonly mutated chromatin modifier in MB as a whole, MLL2, to elevate H3K4me3 and remove H3K27me3 marks in order to activate target genes 65, 66. Robinson et al further speculated that this loss of KDM6A function, as well as additional mutations in ZMYM3 and CHD7 (regulators of H3K4me3‐mediated patterns) and overexpression of the H3K27 methyltransferase EZH2, may lead to an imbalance between trimethylated H3K4 and H3K27 and altered differentiation signaling in this subgroup 62 (see Figure 2).

Figure 2.

Schematic of key chromatin‐related mutations in medulloblastoma (MB). This schematic view indicates some of the recurrently mutated chromatin modifier genes in medulloblastoma, and their possible effects on disrupting chromatin marks, with a particular focus on the disrupted H3K4/H3K27 methylation frequently observed in Group 3 and Group 4 tumors.

In a follow‐up study of these sequencing efforts, Dubuc et al screened a large series of MBs specifically for mutations in MLL2 and KDM6A 17. MLL2 mutations were seen in approximately 8% of MBs, with roughly half of those predicted to be truncating alterations and no clear enrichment in a particular MB subgroup. Alterations of KDM6A, in contrast, were largely restricted to Group 4 tumors (in keeping with the above sequencing and copy‐number studies) and were mostly truncating 17. The two alterations were found to be mutually exclusive, as might be expected from the above‐noted similarities in the predicted outcome of these changes. When further considering copy‐number and transcriptional changes of EZH2, KDM6A and KDM6B, the authors found further support for the concept postulated by Robinson et al 62 that a subset of Group 3 and Group 4 MBs may show an enrichment of H3K27me3 marks (“K27+”). This was further supported by immunohistochemical staining for various histone marks, as well as an overrepresentation of differential expression of polycomb repressive complex 2 (PRC2) targets (which are typically linked to H3K27me3 levels) in Group 3 and Group 4 tumors with or without this K27+ phenotype 17.

As with copy‐number changes, evidence has recently emerged that the relationship between these DNA mutations and chromatin architecture may not be solely a one‐way process. For example, it seems that certain marks of heterochromatin, particularly H3K9 trimethylation, are highly enriched in regions of the genome that show a higher rate of mutation 67, indicating perhaps that DNA repair processes are not as efficient in these repressive chromatin domains. It is unclear as yet whether a similar phenomenon may be shaping the mutational landscape of somatic changes in MB.

Other Links Between Chromatin Alterations and MB

As noted earlier, several of the reported alterations in chromatin modifying genes seem to converge on methylation of lysine 27 of histone 3, particularly elevated H3K27me3 in Group 3 and Group 4 tumors 17, 62. In a similar vein, one recent report investigated the link between polycomb group genes (required for induction of H3K27 methylation) and the transcription factor OTX2 7, which is occasionally amplified and almost universally overexpressed in Group 3 and Group 4 tumors 1, 6, 50. The authors found upregulation of multiple genes belonging to PRC1 and PRC2, including EZH2, EED and SUZ12, across MB subgroups, but particularly in Group 3 and Group 4 MBs 7. Upon silencing of OTX2 in the MB cell line D425, Bunt et al saw downregulation of these polycomb genes in conjunction with induction of H3K27 demethylases (e.g. KDM6A, KDM6B and KDM7A), resulting in reduced levels of H3K27me3 7. Thus, further investigation of the role of OTX2 as a regulator of key chromatin modifiers in certain MB subgroups may be warranted.

In addition to these OTX2‐mediated changes, other examples of polycomb deregulation have also been reported. In particular, the PRC1 regulator and stem cell self‐renewal factor BMI1 has been linked to SHH‐MB [the only MB subgroup in which OTX2 is essentially not expressed 1]. Using Bmi1‐null mice, Leung et al first showed in 2004 that Bmi1 is essential for normal cerebellar development 40. The same study identified a subset of human MB characterized by concordant overexpression of PTCH1 and BMI1. This was built upon by Michael et al in 2008, in a study which used a SmoA1 mouse model to demonstrate that Bmi1 is essential for SHH‐MB tumorigenesis 43. A recent report also identified a feedback loop between hedgehog pathway activation and BMI1, whereby Shh ligand can induce expression of BMI1 in MB tumor‐initiating cells, which then further upregulates hedgehog target genes 79.

Outlook

The current era of MB research is an exciting one. Significant advances have been made over the past few years, with the development of important subgroup classification schema and the identification of multiple prognostic markers. As large‐scale genomics projects have come to fruition, chromatin remodeling defects have taken center stage in multiple studies. There is still much to be done, however, in translating this increased knowledge into true clinical benefit. While the genetic alterations affecting chromatin remodelers are becoming better understood, the downstream epigenetic consequences of these alterations remain relatively understudied. In addition, it is not clear whether these changes in chromatin architecture are really seen across the whole genome, or whether they affect only a subset of specific targets. The transcriptional consequences of mutations in chromatin modifiers are also still somewhat unclear. More detailed characterization of the MB epigenome will therefore be an important area of focus in the coming years. Several such studies are under way, for example, under the auspices of International Cancer Genome Consortium (ICGC) projects in Germany and Canada or the Pediatric Cancer Genome Project in the USA, to look at global DNA methylation (Illumina Infinium 450k arrays and whole‐genome bisulfite sequencing), non‐coding RNAs (RNA‐seq, miRNA‐seq) and patterns of critical histone modifications (ChIP‐seq).

The repertoire of potential therapeutic targets has also expanded rapidly in the genomics era. Many novel therapeutic agents targeting chromatin modifiers are currently in development or in early stage trials, although not many have achieved Food and Drug Administration (FDA) approval as yet 5. The best known class of approved drugs are histone deacetylase inhibitors such as vorinostat 31 and romidepsin 26, which are both in phase I/II clinical trials for pediatric brain tumors (see http://www.clinicaltrials.gov). Vorinostat has shown some promising results in preclinical MB models 44, 46, 69, 74, and knockdown of specific HDACs in MB cells results in decreased proliferation and cell viability 45. More recently, inhibition of bromodomain‐containing proteins, which interact with acetylated histones and recruit transcriptional regulators 59, has been suggested as one way of indirectly targeting Myc‐driven tumorigenesis 15. This is of particular interest given the strong transcriptional upregulation and/or amplification of c‐Myc in WNT and Group 3 MBs. Targeting the aberrant regulation of H3K27 methylation observed in MB has also been suggested as a therapeutic avenue 17, 62. As further support for this concept, DZNep (3‐deazaneplanocin A), an inhibitor of the H3K27 methyltransferase EZH2, has recently been shown to suppress MB cell growth in vitro 2. Phase II clinical trials of EZH2 inhibitors are now part of the Children's Oncology Group (COG) blueprint for MB research 20. Particularly for those tumors with elevated OTX2 expression, the combination of polycomb inhibition with a differentiating agent such as all‐trans retinoic acid (ATRA) has also been suggested 4, 7, 16. As it seems, however, that the global process of histone methylation is affected, rather than a single gene or pathway, therapeutic targeting of these alterations may prove challenging.

Finally, the development of tumor models that faithfully recapitulate these epigenetic alterations will also be crucial, both for further characterizing the biological changes resulting from epigenetic dysregulation and for the development and testing of novel therapeutics. The majority of the established MB cell lines are thought to most closely resemble Group 3 MB, with no WNT models and only poor representation of the SHH and Group 4 subgroups. As with any long‐term culture model, there are also concerns as to how well these lines truly reflect the primary tumor. As such, a lot of effort is currently being put into developing suitable in vivo models. There are already genetically engineered mouse models (GEMMs) of three of the four molecular subgroups of MB 18: WNT 21, SHH 22, 24, 27, 80, 82, 83 and Group 3 32, 54. A MYCN‐driven model that may resemble Group 4 tumors in certain contexts has also been established 71. None of these GEMMs, however, are driven by alterations in chromatin modifiers, and it will therefore be necessary to build a larger repertoire to model the variety of changes identified through high‐throughput genomics. A number of research groups are also building collections of xenografted tumor material from the various molecular subgroups, which will help recapitulate the wide spectrum of alterations observed in primary tumors and provide a useful additional tool for preclinical testing of targeted agents 85.

Thus, while it is right to acknowledge the multiple advances that have recently been made in the field, building on these insights into the role of chromatin remodeling in MB in order to transfer this knowledge to the bedside will be one of the major challenges in the emerging post‐genomics era.